What is the expected use-case for resize! in the context of DDE problems?
I would have expected that many users like myself would use it to introduce new dependent variables into the system. For many DDE models this new variable will be dependent on its own history. Is it expected that users initialise a history function with large enough dimension to accommodate all possible dependent variables in the model, prior to solving?
Or is it better practice to avoid use of the resize! function and iteratively solve for intervals of time, re-defining the system and history as we go based on sol(t) e.g.
sol # solution of previous time interval
global n = length(sol.u[end]) # size of system in previous time interval
global init # initial condition of my new dependent variable
function h(p, t)
output = Vector{Float64}(undef, n + 1) # re-size system by one
for i = 1:n
output[i] = sol(t)[i]
end
output[n + 1] = init
output
end
# define a new u0 with length n + 1, solve, etc.
The latter approach seems more intuitive to me. I bring this up because I think the purpose of resize! might be confusing to users, if they have to initialise the history function for all future variables a priori.